The advertising business is waiting for a revolution

An impressive demonstration of the work of 2 neural networks — Midjourney and Kling AI. The first generates an image, the second animates it according to your arrows, and the output is a professional video literally out of nothing.

A cheat sheet of useful applications equipped with artificial intelligence

Save it 👨‍💻

Conversational AI 🤖

Create artificial intelligence-based agents that can speak in minutes, with low latency and full customizability.

A huge database of neural networks — ChatGPT, Claude and other LLMs:

• AI collections for any task;
• Courses from top universities and a list of scientific articles;
• YouTube channels with the analysis of neurons.

Gifts from the Mask 🎁

Twitter gave everyone the Grok neural network — the most crazy chatbot became available to everyone for free.

The first uncensored AI works directly on the social network X. He knows how to joke, knows the latest news, answers the most pressing questions and generates "rebellious" pictures. The only drawback is that in the free version, only 10 text requests are available every 2 hours and 3 pictures per day.

❗️Google Announces Gemini 2.0: A New Era of AI Agents

What's New in Gemini 2.0?

The new model brings groundbreaking capabilities to the table. Unlike its predecessor, Gemini 2.0 can not only understand multiple types of input but also generate various forms of output. It can create images, produce audio, and seamlessly integrate with tools like Google Search.

3 Groundbreaking Projects

Google is showcasing Gemini 2.0's capabilities through three innovative projects:

1. Project Astra: An enhanced universal AI assistant that can understand multiple languages, remember conversations for up to 10 minutes, and naturally integrate with Google tools like Search, Lens, and Maps.

2. Project Mariner. A browser-based AI agent that can navigate web interfaces and complete complex tasks.

3. Jules: An AI-powered coding assistant that integrates with GitHub, helping developers plan and execute coding tasks under their supervision.

AI tokens are considered a promising investment category in 2025

Collectively, the FET, NEAR, RENDER, TAO, ROSE and AR tokens have a market capitalization of over $44.5 billion.

🟦FET. After the correction in July, by September 2025, analysts predict an increase in the crypto asset to $ 30.

🟦NEAR. The token shows steady growth with the possibility of reaching $10 in December.

🟦RENDER. The growth potential depends on holding above the $8.20 support level.

🟦TAO. The token is currently consolidating around $650, with new potential historical highs projected at $12,000.

🟦ROSE. The recent consolidation phase is seen as a buying opportunity on the decline near the $0.11 level.

🟦AR. The augmented reality token recorded the highest trading volume in November 2024, and targets are now set at $86.51.

The AI ​​Training Revolution: From Humans to AI. Why Amazon, Meta, and Google Are Investing Billions Against Expectations

While critics point to benchmark saturation and data exhaustion, the biggest companies are showing confidence with massive investments.

The Infrastructure Race:

• Amazon: Investing in 400,000 Trainium2 Chips for Anthropic

• Meta: Building a 2GW Data Center in Louisiana

• OpenAI and Google: Developing Multi-Data Center Training

The AI ​​Training Revolution: From Humans to AI.

Old approach: RLHF (Meta* spent $20M on Llama 2)
New approach: RLAIF (feedback from AI).

Constitutional AI by Anthropic introduces a new approach to creating ethical models with a two-stage learning system.

Companies' solutions:

- OpenAI: inference-time optimization
- Anthropic: two-stage model (Opus/Sonnet)
- Meta: efficient learning algorithms.

Future trends:

❗️Video data processing as a new source of knowledge
❗️Multi-data-center learning
❗️Reasoning capacity enhancement
❗️Feedback automation

Google announced Veo 2: a SOTA video generation model which produces realistic, high-quality clips from text or image prompts.

They’re also released an improved version of text-to-image model, Imagen 3 - available to use in ImageFX through.

The AI Video Composer neural network is your best friend in editing

It works simply: we upload images / video / audio, and then we write the promt what exactly needs to be done. The service will do everything itself in a couple of minutes, leaving only the stage of making edits.

Noetik has created an AI that simulates the behavior of cells in human tissues

OCTO-VirtualCell can predict how different cells in a particular patient's tissues will behave.

Imagine that you can "place" a virtual cell anywhere in the tumor and see how it will react — that's exactly what this system does.

To create this technology, Noetik used:
- Data of more than 1000 patients
- 40 million cells
- One of the largest datasets of this type in the world.

This development opens up new opportunities for:

- Personalized medicine:

- Development of new drugs:

- Study of various diseases:

In the updated Canvas in ChatGPT, you can quickly translate code from one programming language to another — and it doesn't matter which one, your own or generated

Instagram is preparing to launch its AI video editor.

Adam Mosseri, the head of the company, announced the Movie Gen AI AI tool, which will allow editing videos based on text queries.

Voice AI Market Map: 2024 Results and 2025 Forecasts, Cartesia Report

The cost of using language models has fallen dramatically: from $45 to $2.75 per million tokens. At the same time, the quality of speech recognition and synthesis has increased significantly.

Interest in voice technologies is growing rapidly - the number of startups in this field at Y Combinator increased by 70% between the winter and fall 2024 intakes. Voice AI assistants are actively being implemented in:
- Healthcare:
- Insurance:
- Logistics:
- Hotel business:
- Small business:

What awaits us in 2025?

1. More advanced speech-to-speech conversion systems with a delay of only 160 ms are expected to appear (for comparison: a person has 230 ms).

2. The development of compact models will allow voice assistants to be used without the Internet - on phones, in cars and various devices.

3. Voice assistants will begin to cope with complex tasks.

2025 promises to be the year of mass implementation of voice technologies.